RDP 2023-02: Did Labour Market Concentration Lower Wages Growth Pre-COVID? 6. Why has the impact of concentration increased?

A key question from a policy standpoint is, why has the impact of concentration increased? Understanding this is crucial in identifying what, if any, polices could and should be implemented to offset the apparent increase in market power.

I examine three potential explanations motivated by the existing literature on firm dynamism and labour market concentration:

  • Declining industry or occupational mobility
  • Declining firm dynamism and entry
  • Declining union membership.

6.1. Declining occupational mobility

One explanation I examine is declining occupational or industry mobility. Much of the work on labour market concentration assumes that workers cannot move between local labour markets due to geographic or skill barriers. However, in practice workers can move between geographies, industries or occupations, even if it is costly or difficult. If it becomes easier to move between these markets, and so the set of potential outside options becomes larger, the effect of concentration within the market might become smaller.

For example, in the Berger et al (2021) model they allow for differentiation in supply of labour across, as well as within industries. When differentiation across industries declines, consistent with easier mobility across industries, the impact of differentiation and concentration within industries is lessened.

Schubert et al (2021) focus on this issue empirically. They derive a model where workers can potentially move to other industries in their locality. They show that in this case wages will be impacted by the ‘outside options’ that other occupations in their locality represent. Failing to include these outside options can bias the estimated impact of the HHI on wages. Similarly, they show that the impact of labour market concentration on wages tends to be smaller in industries with more cross-industry mobility, and the outside options become more important.

Declining occupational mobility and increasing frictions in moving between industries and occupations may be relevant in the Australian case. For example, Figure 8 shows that a moderate portion of the decline in job-to-job switching (as defined in Deutscher 2019) reflects fewer people changing occupations. This might suggest increasing frictions between labour markets.

Figure 8: Job-to-job switching by type
Figure 8: Job-to-job switching by type - This chart shows the rate of job switching over time, and decomposes it in the portion where the person stays in the same location and occupation (averages 11 per cent), switches occupation (averages around 5 per cent), or switches location (averages around 1 per cent). The overall rate of switching declined from around 16 per cent in the 2000s, to around 14 per cent in the 2010s. Much of the decline reflects less people switching occupations, which fell from around 5 to 3 per cent. Location switching fell slightly too from 2013.

Note: Job-to-job switching as defined in Deutscher (2019). Location defined as SA4 or greater capital city area. Occupation is 2-digit ANZSCO.

Source: Author's calculation

To consider this, I include the measure of outside options suggested by Schubert et al (2021) and examine the impact on the coefficient on the HHI in the full sample, and in the sub-sample regressions.

The outside-option variable is constructed as a weighted average of the wages for firms in all other industries in the local area. The weights are a function of two factors: the share of employment for each of these industries in the local area (excluding the base industry), to capture how ‘important’ each industry is in the local area; and the national rate of switching between the base and the other industry (measured as a share of job-switchers only), to capture the ease with which people can move from the base industry to the alternative industry.[11]

The results of the regression are contained in Table 5. Focusing on Column 2, we can see that having strong outside options leads to higher wages. Moreover, the effect of concentration within a market is lower in this regression, suggesting that failing to account for these options can cause us to overstate the effects of the HHI.

Focusing on the sub-sample in Column 3, we see outside options becoming less impactful over time. This is consistent with what we might expect to see if it became harder to move between industries. This suggests that declining occupational mobility may have weighed on wages growth by limiting workers' ability to leverage options outside their industries in wage negotiations, which would be consistent the findings in Deutscher (2019) that areas with less job switching tend to see lower wages growth.

Table 5: Outside option specification
  Baseline With OO Sub-sample OO
Concentration −0.122***
Outside option   0.040***
Concentration * GFC     −0.076***
Concentration * Post GFC     −0.083***
Outside option * GFC     −0.010***
Outside option * Post GFC     −0.020***
R2 0.670 0.671 0.671
N 302,000 302,000 302,000
Fixed effects
Local market Y Y Y
Time*Industry Y Y Y
Time*Location Y Y Y
Note: All regressions include a set of common controls covering market characteristics (number of workers in the market). All controls other than firm FE are interacted with period dummies. Pre-GFC is 2005–2007. GFC is 2008–2010. Post GFC is 2011 to 2015. Errors clustered at the local market level.

Still, this regression provides no evidence that changes in the importance of outside options and occupational mobility have increased the impact of concentration. The coefficient on the HHI declines by a similar amount over the sample as in the model without outside options, so changes in the importance of the omitted outside options variable was not driving the decline.

To test more directly the impact of declining occupational mobility on the role of concentration, I extend the baseline market-level monopsony model to incorporate industry switching rates. Specifically:

ln( w i,t )=γ*Pro d i,t + X i,t +β*ln( Conce n i,t )+δ*ln( Conce n i,t )*Switc h ind,t +θ*ln( O O i,t )+ρ*ln( O O i,t )*Switc h ind,t + i,t

Switchind,it is the national rate of industry switching in the relevant ANZSIC 3-digit industry. This is either expressed as a share of all job switchers (leave share), or as a share of all job switchers and stayers (leave rate) with both defined as in Deutscher (2019) and focusing on worker's main jobs.

I focus on the national, rather than local rate of switching, to avoid endogeneity between local conditions and switching rates. It also helps to limit noise introduced by very small markets, where rates can be very volatile. Switch rates are demeaned for each industry. As such we are focusing on changes in switching rates and abstracting from structural differences across industries that might affect switching, concentration and wages.

The coefficients of interest in this case are δ and ρ . If concentration amongst incumbents is less impactful in industries with more occupational switching, the coefficient δ would be positive. Similarly, if outside options are more important where switching is more prevalent ρ would be positive.

Table 6: Leave share, concentration and wages specification
  OO model OO model with leave share OO model with leave rate
Concentration −0.104***
Outside option 0.040***
Concentration * Leave share   −0.288*
Outside option * Leave share   0.086***
Concentration * Leave rate     0.035
Outside option * Leave rate     0.228***
R2 0.671 0.671 0.671
N 302,000 299,000 302,000
Fixed effects
Local market Y Y Y
Time*Industry Y Y Y
Time*Location Y Y Y
Note: All regressions include a set of common controls covering market characteristics (number of workers in the market). Leave share captures share of job leavers that change industry. Leave rate captures share of total workers in the industry (leavers and stayers) that change industry. Errors clustered at the local market level.

Using both the leave share and rate, we see no evidence that concentration is less impactful where there is more occupational switching (Table 6). However, there is some evidence that outside options are more important in industries with higher switching rates.

Taken together, these results provide some evidence that decreasing occupation and industry mobility may have weighed on wages pre-COVID by limiting the role of outside options. However, the direct evidence with respect to its role in increasing the impact of concentration is limited.

6.2. Declining firm dynamism

The role of firm entry and dynamism in labour market dynamism has been considered in some recent papers (Bilal et al 2019). In these papers, firms enter and attract staff away from incumbent firms, meaning that declines in firm entry can be associated with lower job-to-job switching and few outside options for workers.

Such dynamics are evident in Australia. Much of the decline in job-to-job switching, as defined in Deutscher (2019), has reflected less workers switching from established mature firms to young firms. And this has coincided with declines in firm entry rates (Figure 9).

These patterns are also evident within local areas. Figure 10 shows that there is a strong positive correlation between firm entry rates in a local area and the rate of job-to-job switching. This is the case even controlling for time-invariant differences across geographies and the overall business cycle. Moreover, this relationship is only evident when focusing on switching between mature and young firms. There is no relationship between firm entry and mobility of workers between mature incumbent firms, indicating that the relationship is not simply driven by local economic conditions (Appendix A).

Figure 9: Firm entry rates and rate of job-switching from old to young firms
Figure 9: Firm entry rates and rate of job-switching from old to young firms - This chart shows bars for the average rate of job switching, and the rates of switching between firms of different ages. It splits these into bars for the period 2005-2008 and 2012-2016. The total switching rate fell from around 16 per cent to 14 per cent. Switching from young to mature, and young to young firms fell very slightly, after being around 2 per cent in the early period. Switching from mature to mature fell slightly from around 9 per cent. And switching from mature to young firms fell more substantially from around 4 to 2 per cent. Over the same period, firm entry rates fell from around 14 per cent to 11 per cent.
Figure 10: Entry rate vs job-to-job switching by SA4 for 2002–2016
Figure 10: Entry rate vs job-to-job switching by SA4 for 2002–2016 - This chart is a scatter plot of entry rates for areas, against job switching rates in the area. There is a strong positive relationship, with switching rates increasing as entry rates increase.

Figure 9 young firms are 0–5, mature firms are older than 5 years. Figure 10 plots entry rates for SA4s or greater capital city areas versus the job-to-job switching rate. This is shown having residualised with respect to time effects, area effects, and the employment share of young firms. Approximately 780 SA4*year combinations based on 120 million workers.

Source: Author's calculations

Taken together, these results suggest that firm entry represents an additional source of competition for labour for incumbents. Entrants could broaden the set of outside option for workers, and so raise their bargaining power. As such, declining firm entry and dynamism could make concentration among incumbents more impactful.

To examine the role of declining firm entry, I extend the baseline market-level monopsony model to incorporate firm entry rates. Specifically:

ln( w i,t )=γ*Pro d i,t + X i,t +β*ln( Conce n i,t )+δ*ln( Conce n i,t )*Entr y ind,t + i,t

Entryind,t is the national rate of entry in the relevant ANZSIC 3-digit industry. I again focus on the national, rather than local rate of entry and use demeaned entry rates.

The coefficient of interest in this case is δ . If concentration among incumbents is less impactful in industries with rising entry, the coefficient would be positive.

Table 7 shows the results of the regression with entry rates constructed using the entries of firms, and entries of plants (recording a firm opening a plant in a new locality as an entry). The results consistently show that, where industry entry rates are higher, local labour market concentration is less impactful.[12] That is, having a high level of firm entry seems to provide an additional source of competition, lessening the market power associated with concentration among incumbents.

Table 7: Firm entry, concentration and wages specification
  Baseline Plant-based entry rate Firm-based entry rate
Concentration (HHI) −0.122***
Concentration (HHI) * Entry Rate   0.540***
R2 0.67 0.683 0.686
N 302,000 274,000 274,000
Fixed effects
Local market Y Y Y
Time*Industry Y Y Y
Time*Location Y Y Y
Note: All regressions include a set of common controls covering market characteristics (number of workers in the market. Errors clustered at the local market level.

To provide a sense of the macro importance of declining entry rates I do a similar calculation as for the sub-sample regressions. On average there was around a 5 percentage point decline in the entry rate of plants and firms across the sample. Applying this to the coefficients from Columns 2 and 3 of Table 6 and using the employment-weighted average HHI (0.1), this suggests that wages would have been around 0.250.50 per cent higher in 2015, had entry rates not declined.

Again, the estimate should not be interpreted too precisely as it is a simple, partial equilibrium estimate. But these results do suggest that declining firm entry and dynamism had a substantial negative impact on workers' wages by lowering their bargaining power and increasing the power of incumbent firms.

6.3. Declining union membership

Benmelech et al (2022) argue that declining union membership rates could account for the increased impact of concentration on wages in the US. They argue that unions provide an offset to monopsony power. With lower union membership there is less of an offset, and so concentration could be more impactful.

Union membership rates in Australia have been declining for decades. And while Bishop and Chan (2019) found no evidence that declining union membership led to lower wage outcomes in Enterprise Bargaining agreements in Australia – as the share of agreements with union involvement remained unchanged – it is still worth examining the interaction between union coverage, labour market concentration and wages.

To do so, I run a similar specification to the above, and to that used in Benmelech et al (2022):

w i,t =γ*Pro d i,t + X i,t +β*Conce n i,t +δ*Conce n i,t *Unionshar e div,t + i,t

Union sharediv,t is the share of employees in each ANZSIC division that are part of a union. This is taken from the ABS Employee Earning, Benefit and Trade Union Membership release for 2006–2013. Unfortunately, more granular industry-level data were not available. Again, rates were demeaned to focus on changes in union membership, rather than the levels.

Table 8 shows the results. There is a significant positive coefficient on δ , indicating that the impact of market concentration on wages is smaller (in absolute terms) when union membership rates are higher. This does suggest that declining unionisation rates may explain part of the increased impact of concentration on wages. However, declining entry rates appear to be the more important factor, particularly once we try to account for both entry and union coverage (Column 3).

More generally, it is important to keep in mind that these models only capture one mechanism through which union coverage can affect wages. They do not account for any other effects changes in union membership rates could have on the economy, in terms of productivity, employment or output, and do not speak more broadly to the net benefits or costs of lower union membership. They simply indicate that, in the presence of monopsony power stemming from concentrated labour markets, collective worker bargaining can play an offset, and that this offset may have declined over time.

Table 8: Union membership, concentration and wages specification
  Baseline Union Union and Entry
Concentration −0.133***
Concentration * Entry Rate     0.939***
Concentration * Union   0.009**
R2 0.67 0.70 0.70
N 247,000 219,000 219,000
Fixed effects
Local market Y Y Y
Time*Industry Y Y Y
Time*Location Y Y Y
Note: All regressions include a set of common controls covering market characteristics (number of workers in the market and market fixed effects). Regressions 2006–2013. Errors clustered at the local market level.

One related explanation not explored in detail due to data constraints is changes in the share of workers on different pay-setting mechanisms. To the extent that pay is set nationally for a firm or industry, local concentration may be less impactful. This does not necessarily mean that wages will be higher, but the relationship between local concentration and wages may be weaker.

Unfortunately, data on pay-setting shares by ANZSIC division are only available for three years across my sample: 2008, 2012 and 2014. Based on this limited data, there is cross-sectional evidence that the HHI is more impactful in divisions with higher use of individual agreements. However, it is possible that this is driven by some common factor affecting both, such as level of education required for staff. Focusing within sectors, there is no evidence that the HHI has become relatively more impactful where the individual agreements usage increased.[13]

With additional data this channel could be assessed more directly. However, given wages in individual agreements have to be at least as high as award wages, and given that individual agreement usage increased up to 2008, but then fell, increased use of individual awards seems unlikely to be behind the results in this paper.


Note that due to the inclusion of location by time-fixed effects in the regression, we are identifying whether wages increase more in industries where wages in closely linked industries rose relatively more. This makes it difficult to use this set-up to understand macroeconomic implications, rather than simply examining the mechanism. For example, taken at face value, if the outside option becomes less important, this could help workers with poor outside options. [11]

These results are robust to using different methods for calculating entry rates that are more robust to outliers, and using measures of churn, rather than entry (results available on request). [12]

Results available on request. [13]